WEBVTT 00:00:00.000 --> 00:00:10.000 Alright, thanks, so much, Jean, and thanks everybody for the opportunity to present here. We're really excited to talk to everyone just by way of quick introduction. 00:00:10.000 --> 00:00:22.000 So, just by way of quick introduction. So Jim Hawka unfortunately, could not be here today for this presentation she is a statistician at Pnnl, and leads a lot of the this work that I work on my name is Lisa Newburn. 00:00:22.000 --> 00:00:22.000 I've been a piano for about 19 years. 00:00:22.000 --> 00:00:31.000 I primarily work on the visual sample plan software, which I will talk a little bit about today and I'm just generally interested in developing software tools for statistical sampling design. 00:00:31.000 --> 00:00:40.000 Fred, would you like to give yourself a quick introduction? 00:00:40.000 --> 00:00:42.000 Thanks, Lisa. My name is Friday. Lewis. 00:00:42.000 --> 00:00:55.000 I've been at Pnnl for about 2 years, and I work in near surface geophysics and hydrogeophysics, and my background is in hydro geology and geophysics. 00:00:55.000 --> 00:01:14.000 Alright great! And so with that we will get started. So for our presentation today, we're going to be talking about some of the work that's happening at piano and some of what we've been doing in terms of data analysis and statistical sampling for site characterization so here the methods for 00:01:14.000 --> 00:01:18.000 statistical sampling and analysis for later phases of site remediation. 00:01:18.000 --> 00:01:35.000 Often well established where there's regulations and guidance that specify the methods to be used, and the data collection is often really driven by these specific statistical requirements, whereas in the characterization phase not only are there a whole lot of different sources of data 00:01:33.000 --> 00:01:55.000 A continually evolving conceptual site model. The types of analysis can often be fairly subjective and qualitative and not be driven by these kind of external criteria or guidance guidance, you know, specific specific requirements, and so some of the work. 00:01:55.000 --> 00:02:10.000 We're doing is looking at ways that systematic planning data quality assurance, ge, physical data, analysis and statistical sampling design can assist in site characterization and we're not claiming to have all the answers of the question of how best to do this how how best to use data analysis and 00:02:10.000 --> 00:02:18.000 statistical sampling design to address this problem. But we're working to answer some of those questions. 00:02:18.000 --> 00:02:29.000 So here is an example from the Icrc website of a systematic process for doing site characterization for doing integrated site, characterization and kind of doing continual refinement of a conceptual site model. 00:02:29.000 --> 00:02:40.000 So the initial steps here in the blue talk about kind of similar to a data quality objectives process where you're really defining the problem. 00:02:40.000 --> 00:02:55.000 And the uncertainties identifying the data needs and gaps identifying the objectives for data collection that you're trying to collect data to address and developing the process. 00:02:55.000 --> 00:03:01.000 The next section there talks about or illustrates these convictive tools that might be useful. 00:03:01.000 --> 00:03:20.000 As part of this process. And then finally, once the investigation is implemented and executed, the conceptual site model can be updated and evaluated to determine if those objectives were met and then kind of the main thing wanted to talk about here is just the fact that this really is likely to be an 00:03:20.000 --> 00:03:24.000 iterative process, where data gathered will feedback and do this continual refinement. 00:03:24.000 --> 00:03:30.000 And so, particularly when dealing with a site that's very complex for us, where we're remediation could take a long time. 00:03:30.000 --> 00:03:40.000 An approach like this could be warranted. And so, where the items we're gonna talk about today, where they fit in. 00:03:40.000 --> 00:03:45.000 So I'll be talking a little bit about the visual sample plan software. 00:03:45.000 --> 00:03:56.000 Vsp, that can kind of fit into the process by helping to design the data collection and analysis process, and then also in performing the data, evaluation and interpretation. 00:03:56.000 --> 00:04:05.000 Especially in bringing in multiple lines of evidence. And then the geophysics. 00:04:05.000 --> 00:04:18.000 Part of our talk is talking about. This is another type of investigative tool that can be part of this toolbox for performing site characterization. 00:04:18.000 --> 00:04:25.000 Just by way of example. Really, some of the data that might be available for survey planning or available. 00:04:25.000 --> 00:04:35.000 You know the characterization process is really just a huge huge array of data that that may be coming in and not intending to read through all this. 00:04:35.000 --> 00:04:39.000 But there's gonna be hundreds to thousands of different types of data coming in through different phases. 00:04:39.000 --> 00:04:52.000 So part of what we're considering is how some of the common data types can be used in planning and statistical analysis, in characterization and in subsequent phases. 00:04:52.000 --> 00:05:11.000 This diagram here. Many people might be familiar with that kind of the data quality assurance process, and how the quality assurance, the data quality assurance comes in in the assessment phase where it's applied to the data collected prior to and from the characterization survey so historically, this is kind 00:05:11.000 --> 00:05:27.000 of called out as a requirement during the assessment phase. And we in this paradigm, we're wanting to enable the user to use that previous data from various sources and make Dqa really part of the planning process as well. 00:05:27.000 --> 00:05:37.000 So to start off with, I will turn it over to Fred to talk about data analysis for characterization in particular, geophysical characteristics. 00:05:37.000 --> 00:05:37.000 So, Fred, I believe you could have click control on this. 00:05:37.000 --> 00:05:46.000 Are you able to? Are you able to move through the slides? 00:05:46.000 --> 00:05:49.000 I let's see, I'm trying. 00:05:49.000 --> 00:05:53.000 If not, I'm happy to advance the slides. 00:05:53.000 --> 00:05:55.000 I think we need you to advance the slides. Lisa. 00:05:55.000 --> 00:05:57.000 No, problem. Just let me know when. 00:05:57.000 --> 00:06:15.000 So next slide, please. Great, thank you. So, geophysical methods are used to fill gaps in space between sparse measurements taken at monitoring or sampling points and potentially also to fill gaps in time monitoring isn't the focus of our talk today, but there are increasing 00:06:15.000 --> 00:06:19.000 applications of geophysics in time, lapse mode at the right. 00:06:19.000 --> 00:06:28.000 We have a schematic that makes the analogy between medical and geophysical imaging, as everyone knows, over the last few decades, cat scans Mris and other imaging technologies. 00:06:28.000 --> 00:06:50.000 Have revolutionized diagnostic medicine, allowing for non-invasive detectection and time lapse or functional monitoring of flow and transport processes within the human body, similarly, geophysical imaging has potential to transform how we characterize and also monitor environmental 00:06:50.000 --> 00:06:59.000 sites, filling gaps in space and time in doing so, by not invasive or minimally invasive means. 00:06:59.000 --> 00:07:03.000 No physical imaging can be performed in 2D. Or 3D. 00:07:03.000 --> 00:07:19.000 Perspectively to yield slices or volumes of sub-surface properties, such as electrical conductivity or it's reciprocal resistivity seismic velocity, radar velocity, or other physical properties in many cases, these geophysical 00:07:18.000 --> 00:07:29.000 Solids, Water Saturation, and geologic features in the subsurface. 00:07:29.000 --> 00:07:36.000 So they provide information that we can use to refine conceptual site models. 00:07:36.000 --> 00:07:53.000 And Hi hydro geologic frameworks, and also to reduce the uncertainty associated with those models. 00:07:43.000 --> 00:08:00.000 It's important to note, however, that these different geophysical properties I'm talking about are only indirectly related to the properties that we would really want to update our Csms with rare exceptions, geophysical measurements don't directly tell us about permeability for 00:08:00.000 --> 00:08:05.000 Usity, concentration, saturation, rock type, etc. 00:08:05.000 --> 00:08:11.000 And there's an interpretive step involved in making sense of the geophysical results. 00:08:11.000 --> 00:08:21.000 In the second schematic at the right is an example of resistance, activity, imaging for characterization of hydrostatigraphy. 00:08:21.000 --> 00:08:39.000 So here the different resistivity features are interpreted in the context of the local geology to identify a different strata at the site going beyond that sort of conceptual or qualitative interpretation of geophysical images is a topic. 00:08:39.000 --> 00:08:47.000 Of increasing interest, and you have statistical algorithms, some of which are already within Vsp and others planned for implementation. 00:08:47.000 --> 00:09:07.000 Provide a potential framework to achieve this using geophysical information as conditioning data within geostatistical frameworks allows us to integrate sparse direct sampling data with the more spatially exhaustive indirect geophysical information across the site and as we do 00:09:07.000 --> 00:09:17.000 this within the geostatistical framework we can account for the uncertainty associated with the relationship between the geophysical properties and the properties we really care about. 00:09:17.000 --> 00:09:23.000 Again, like permeability, porosity, or concentration. 00:09:23.000 --> 00:09:26.000 Next slide, please. 00:09:26.000 --> 00:09:29.000 Okay. I'm going to walk through a few different examples. 00:09:29.000 --> 00:09:38.000 Demonstrating, geophysics for characterization and the basic idea here is certainly not new to the people who are tuning in today. 00:09:38.000 --> 00:09:52.000 Thanks to the earlier talks, there were several excellent talks, showing examples of geospatics, and talking about different tools for selecting geophysical methods to solve different problems. 00:09:52.000 --> 00:09:53.000 Here, though the examples that I'm showing on this slide are subsurface characterization. 00:09:53.000 --> 00:10:08.000 Using electrical resistivity, tomography, Art can be used to produce snapshots or time lapse, images of electrical resistivity or conductivity. 00:10:08.000 --> 00:10:13.000 Again, these are reciprocal properties in the suburface. 00:10:13.000 --> 00:10:17.000 So we can do this in 2D. Or 3D. I'd left as an example from the Us. 00:10:17.000 --> 00:10:27.000 Department of Energy, Sanford Site, or er T. Results like this have been used to help define hydrography and hydrogeologic framework models. 00:10:27.000 --> 00:10:34.000 So, looking at the spatial distribution of sediment or rock type in the suburface. 00:10:34.000 --> 00:10:34.000 At the right is a second example, using errt at a power plant. 00:10:34.000 --> 00:10:52.000 Site, and the reason that I'm showing this snapshot in particular is just to make the case that er can be used at sites where there is subsurface infrastructure. 00:10:52.000 --> 00:11:11.000 Yes, it's more challenging. But if we know where the underground pipes or tanks or whatnot are located, we can put those into the modeling framework, and I need to give credit for these graphics and the underlying investigations to judy robinson at 00:11:11.000 --> 00:11:24.000 piano L, who's had several papers on the hydrostatography and characterization with Errt and Tim Johnson, who provided the animation of the arc snapshot at the right. 00:11:24.000 --> 00:11:28.000 The primary goal in the work at right was a leak. 00:11:28.000 --> 00:11:42.000 Detection and monitoring. But again, the point of showing this here is just to make the case that we can do this kind of characterization where there is subsurface infrastructure, present. 00:11:42.000 --> 00:11:46.000 Next slide, please. 00:11:46.000 --> 00:11:53.000 This is another example of Ert applied at the Hanford site provided by Tim Johnson. 00:11:53.000 --> 00:11:59.000 This time the investigation was at the Hanford site, B. 00:11:59.000 --> 00:12:18.000 Complex where Vedo stone, contaminant plums associated with leaks or district charges from tanks and trenches, as you see at the right, we've got a subsurface image of electrical conductivity structure and Lisa if you could animate that please. 00:12:10.000 --> 00:12:24.000 Will spin it around so that you can. That's okay, so that we can see the various structures in the sub-surface. 00:12:24.000 --> 00:12:35.000 These plumes derived from tag waste that include radiological and heavy metals, contamination, and these materials manifest a large contrast in electrical properties. 00:12:35.000 --> 00:12:44.000 Compared to the background properties of sediment they don't sound water or groundwater at the site. 00:12:44.000 --> 00:12:51.000 I want to emphasize as I show this, though, that not all contamination has a geophysical signature. 00:12:51.000 --> 00:12:56.000 It depends on the nature of the contamination, and also, of course, contaminant. 00:12:56.000 --> 00:13:08.000 Levels, so for example, you know, part for trillion contamination with pfas would be challenging or impossible to detect in the field using geophysical methods. 00:13:08.000 --> 00:13:18.000 In most cases at least. Here are the presence of contamination increases the electrical conductivity strongly relative to the background, making it a good candidate for er t or other methods. 00:13:18.000 --> 00:13:28.000 Em methods, for example, that can sense contrast and electrical conductivity. 00:13:28.000 --> 00:13:32.000 Next slide. 00:13:32.000 --> 00:13:43.000 Alright in the previous examples we were looking at using geophysical information to refine conceptual site models by way of qualitative interpretation. 00:13:43.000 --> 00:13:49.000 So where's the hydroscopy contacts? 00:13:49.000 --> 00:13:56.000 So where's the contamination in the suburface contributing to various aspects of the Csm. 00:13:56.000 --> 00:14:04.000 This next example demonstrates the use of geostatistical stimulation to quantitatively integrate geophysical data. 00:14:04.000 --> 00:14:09.000 This work was published last year, and a paper LED by Neil Terry, of the Usps. 00:14:09.000 --> 00:14:12.000 And the goal of the work. Our work on this was to map, salinity in the Montebello Basin in California, where water quality was impacted. 00:14:12.000 --> 00:14:36.000 Historically by oil and gas development. We used an indicator geostatistical framework to integrate direct water quality sampling at Limited 4 holes with an extensive database of electromagnetic and other geophysical logs to map salinity distribution throughout the 00:14:36.000 --> 00:14:41.000 basin no, I'm not going to dig too deeply here into the geostatistical methods. 00:14:41.000 --> 00:14:46.000 But the basic idea was to bring together the hard data, the sampling, that is, and the soft data. 00:14:46.000 --> 00:15:02.000 The Em logs capitalizing on the more abundance Geo. Physical information, while accounting for that imperfect relationship between the electromagnetically-drived conductivity and the salinity. 00:15:02.000 --> 00:15:23.000 The framework is probabilistic in nature, so it gives not one best estimate of subsurface salinity, but rather an ensemble of equally likely realizations or models of subsurface salinity, which we can then post process to assess probabilities of exceedance for various thresholds of 00:15:23.000 --> 00:15:30.000 interest, and that's what you see here, Lisa. If you could just click once! 00:15:30.000 --> 00:15:35.000 Will visualize a probability of exceedance for one particular threshold. 00:15:35.000 --> 00:15:49.000 Here a 1,000 milligrams per liter at the site. But we can repeat that for other target thresholds to understand the distribution of salinity in the subsurface. 00:15:49.000 --> 00:16:05.000 So again, these sort of different types of geophysical information are valuable, considering refinement of Csms, and also valuable potentially, quantitatively within geostatistical frameworks. 00:16:05.000 --> 00:16:26.000 So going back to that that Dagram of site, characterization, and the iterative process, we can feed data into the process with geophysics forming informing csms over larger areas inaccessible by wells or other sampling methods and using 00:16:26.000 --> 00:16:42.000 geostatistics understand the uncertainty reduction that we get with the geophysics, and perhaps also to guide further sampling in the future to reduce that uncertainty even more alright. Thank you, Lisa. 00:16:42.000 --> 00:16:51.000 Alright. Thank you, Fred. So now I'll jump back in and talk about methods for characterizing a site with statistical analysis. 00:16:51.000 --> 00:16:52.000 So to start off, I'm just going to give kind of a brief overview of the visual sample plan software. 00:16:52.000 --> 00:16:56.000 So this is a freely available software tool you can get on the Pml website. 00:16:56.000 --> 00:17:03.000 I see that the link is clickable, but for if you can't click it, it's Vsp, P. 00:17:03.000 --> 00:17:23.000 Mnl.gov, and what Vsp is is a software tool that has a few, a few different primary objectives, so the first kind of core one is to help users design a statistically based sampling strategy, namely, how many samples should be taken and where and it's all very based off 00:17:23.000 --> 00:17:36.000 of the data quality objectives process in terms of defining what you want to, what decision you want to collect data to support and making sure that the data you collect is sufficient to meet those objectives. 00:17:36.000 --> 00:17:36.000 Secondly, vsp is performs analysis of data to support decisions. 00:17:36.000 --> 00:17:45.000 So once that data has been collected performing statistical tests. 00:17:45.000 --> 00:17:52.000 Looking at graphs, plots, and summary statistics in order to determine whether the data do support the desired decision. 00:17:52.000 --> 00:17:56.000 Objective. There we go. So vsp, is also the visual part of visual sample plan. 00:17:56.000 --> 00:18:01.000 It's a highly visual tool with a lot of mapping capability. 00:18:01.000 --> 00:18:10.000 Threed building and equipment, visualization we're going to focus on kind of like the twod mapping in this. 00:18:10.000 --> 00:18:23.000 In this presentation, looking at results, and then finally, Vsp is designed to guide users who don't have statistical expertise with the automatically generated report that documents, the assumptions so the Vsp is actually used in a lot of different applications. 00:18:23.000 --> 00:18:35.000 So kind of the traditional use of the Sp or a lot of our user base uses. 00:18:35.000 --> 00:18:43.000 Vsp for environmental characterization and remediation. It's also widely used for decontinination and decommissioning as well as pre-planning. 00:18:43.000 --> 00:18:51.000 For for you know these kind of like long tail events, like as far as biological or chemical, arrayological major incidents, there's a number of different methods in Vsp. 00:18:51.000 --> 00:18:56.000 For groundwater monitoring as well as unexploded ordinance sampling, designing surveys, and doing analysis. 00:18:56.000 --> 00:19:13.000 So basically we you know lot of different areas whenever you need to do sampling to support decisions, there's we've tried to have tools and Psp, that can be that can be used to support that. 00:19:13.000 --> 00:19:23.000 And just briefly, I'm not gonna read all these, but part of, I think, the unique aspect of USP is it has been supported by a lot of different agencies since the ninetys. 00:19:23.000 --> 00:19:29.000 So it's really grown a lot over the years, and it's become a much more powerful tool due to the investments. 00:19:29.000 --> 00:19:36.000 From these different agencies, and combining everything into a into a much more powerful tool that has a lot more capabilities. 00:19:36.000 --> 00:19:56.000 And it is still under active development today. So briefly, just by way of summary, I'm gonna go into some of these in detail in the next slides at like with a detailed example case study in Psp, but just some of the some of the tools within Dsp for looking at data quality are listed 00:19:56.000 --> 00:20:05.000 here just as far as some of the features. And I'm gonna talk about some of these in in in the detailed example coming up here. 00:20:05.000 --> 00:20:12.000 So the scenario that we're kind of looking at here is it's a real scenario. 00:20:12.000 --> 00:20:20.000 Looking at the Hanford BC. Controlled area. So for those who are not familiar, I think you can kind of see the hampered site in my background just to the north of the Pml. 00:20:20.000 --> 00:20:36.000 Campus, but in the mid 1,900 fiftys Hanford retrieved uranium from the oldest tanks of high level waste remaining from the original bismuth phosphate reprocessing process so the uranium recovery worked but generated nearly twice as much waste 00:20:36.000 --> 00:20:52.000 as was originally produced, so the waste for process to scavenge season 137, and strong, 90 to reduce the activities, and then the liquids were discharged to surface oils in a series of covered specific retention trenches so some of can correct me if 00:20:52.000 --> 00:20:56.000 I'm wrong on my picture here, but my understanding is that you had these. 00:20:56.000 --> 00:21:05.000 You had these extractions, plants, and then the waste, the waste, the liquids were were piped over to these, to these trenches over here so that's kinda that's kind of the scenario we're working with. 00:21:05.000 --> 00:21:14.000 Now some of the contamination events following that so the trenches were covered with soil, but we had. 00:21:14.000 --> 00:21:17.000 There are a lot of burrowing native animals out on the site that intruded in, and really really enjoyed that that waste. 00:21:17.000 --> 00:21:25.000 We're kind of using it as like a Salt lake. 00:21:25.000 --> 00:21:27.000 So those Badgers road and chat rabbits. 00:21:27.000 --> 00:21:44.000 And then, of course, those animals spread out, and you know, spread waste, which which spreads contamination further, and then, as we go further up the food chain you get your coyotes and predators who are eating their rabbits and then, further spread their waste across the site, so you're you're 00:21:44.000 --> 00:22:01.000 having all these kind of different pathways of the contamination being spread, and then also kind of famously, for this area these tumbleweeds that have these really really deep roots to get down into where they can reach butter you know kind of sucks things up and and got blown around and 00:22:01.000 --> 00:22:04.000 added to the added to the distribution of contamination. 00:22:04.000 --> 00:22:08.000 So around 1958 to 99 60, this was discovered. 00:22:08.000 --> 00:22:19.000 The holes were filled in and the trenches covered with asphalt, along with additional gravel added, and around 4 square miles was directly impacted and I'm kind of giving a brief summary here if you're interested in a much more deeper dive I really ever recommend this report for better better information. 00:22:19.000 --> 00:22:28.000 Then, and all good. And so here this kind of just shows a map. 00:22:28.000 --> 00:22:35.000 So just for the purposes of this example. What I want to focus on is that we've got this really, really high-level contamination area here. 00:22:35.000 --> 00:22:48.000 And then a known area further out that is expected to be lower contamination or lower levels, but still still definitely defining a contaminated area. 00:22:48.000 --> 00:22:59.000 So for the purposes of of the examples in this presentation, just gonna talk through some possible sampling in analysis, objectives that would take place before remediation, via soil removal. 00:22:59.000 --> 00:23:05.000 And so, of course, all the details, the contamination values in the stamping plans are just for our illustration purposes. 00:23:05.000 --> 00:23:10.000 They've been greatly simplified and modified. 00:23:10.000 --> 00:23:10.000 So starting off. So this kind of shows in Vsp, a couple of decision units that have been delineated. 00:23:10.000 --> 00:23:20.000 So this interior area here, we're calling zone A. 00:23:20.000 --> 00:23:24.000 So this is a region of elevated contamination prior to remediation. 00:23:24.000 --> 00:23:33.000 So this is that really, really elevated area, and then zone B, roughly delineates these areas of where there is elevated contamination. 00:23:33.000 --> 00:23:33.000 But it's just lower risk from this. This zone. 00:23:33.000 --> 00:23:37.000 A so one of the key components of Dsp is, even if you know, we talk about these decision units or sample areas. 00:23:37.000 --> 00:23:49.000 So in later phases, that's often an area that you want to make a particular statistical decision on that. 00:23:49.000 --> 00:23:50.000 In this characterization phase. It often makes sense to divide those up according to that conceptual site model. 00:23:50.000 --> 00:23:55.000 What are the properties of these different areas? What are the ones? 00:23:55.000 --> 00:24:03.000 What areas do we need to treat separately or treat differently because of those properties? 00:24:03.000 --> 00:24:21.000 So, for example, one suppose that prior to remediation, the stakeholders want to estimate the overall season 137 average for both of these songs together, and so they say that this is for planning purposes for the handling and the disposal to remove soil 00:24:21.000 --> 00:24:38.000 and so these 2, these 2 zones are going to differ significantly in the average concentration as well as the standard deviation of the concentration, like the standard deviation of the levels within the area and the size so just roughly, you know zone that zone a was much much smaller than 00:24:36.000 --> 00:24:37.000 And also has a higher standard deviation. It's higher, it's more variability within the area. 00:24:37.000 --> 00:25:01.000 So these are kind of the estimated values based off of some preliminary sampling standard deviation is higher. It's more variability within the area. So these are these are kind of the estimated values based off of based off of some preliminary sampling but the idea. In this example. Is that there needs to be a better estimate 00:25:01.000 --> 00:25:02.000 of so a design. And vsp, that's well suited to this kind of a problem is stratified. 00:25:02.000 --> 00:25:12.000 Sampling. So this is a sampling method that divides a heterogeneous population into these non-overlapping areas or these strata that internally are more harmonious. 00:25:12.000 --> 00:25:15.000 And then you can use different sampling strategies or approaches in each item. 00:25:15.000 --> 00:25:39.000 So some of the advantages of this type of sampling is that it gives you a more accurate estimate of the mean than if you just did simple random sampling across the entire site without stratification, and also allows you to perform better allocation of your samples, hmm so 00:25:39.000 --> 00:25:43.000 stratified sampling can be used for a few different objectives. 00:25:43.000 --> 00:25:46.000 In this case we the objective, is mean estimation. 00:25:46.000 --> 00:25:52.000 So we want to obtain an unbiased, sufficiently precise estimate of the mean and the tool will allocate the samples both by the size of the area and the variability. 00:25:52.000 --> 00:26:01.000 Or the standard deviation in the, in the stratum, and it will. 00:26:01.000 --> 00:26:04.000 But after data on analysis it has. It uses a formula to provide a weighted estimate of both. 00:26:04.000 --> 00:26:13.000 The main and the standard error. So this shows the design dialogue in Vsp. 00:26:13.000 --> 00:26:31.000 Where, based off of the map, the parameters for the Straighta are are loaded in, and then some of the other parameters that estimated standard deviation can also be entered, and the way this works in vsp is you sort of select your optimization method enter in the required 00:26:31.000 --> 00:26:50.000 standard deviation in this case, and then Vsp calculates the total sample size, and then how those samples should be optimally distributed across this across straighta as well as culminating distributed across this across straighta, as well as calculating the total cost so get into a little bit of detail about how the design 00:26:50.000 --> 00:26:53.000 does that optimization? So the different optimization options in this module, you are. 00:26:53.000 --> 00:27:03.000 There's there's 3 different options. So the first one is to minimize the standard deviation of the sample mean for a fixed cost. 00:27:03.000 --> 00:27:10.000 So the idea there is that you're estimating the mean you want to like tighten the error bars on that main as much as you can. 00:27:10.000 --> 00:27:17.000 Given a fixed budget, so I believe that's what that's what I or no, I think I clicked. 00:27:17.000 --> 00:27:26.000 We selected the second one here. Yeah. So the second option here is to minimize the cost for a required standard deviation of the means. 00:27:26.000 --> 00:27:31.000 So that's where you have some specification about how, how those, how tight those error bars must be. 00:27:31.000 --> 00:27:38.000 And you're gonna figure out how many samples and therefore what overall cost is required to achieve that. 00:27:38.000 --> 00:27:43.000 Excuse me, and then finally, the final option here is for a predetermined number. 00:27:43.000 --> 00:27:43.000 So this would be where maybe there's a fixed budget. 00:27:43.000 --> 00:27:47.000 Maybe you know, there's just a certain number of samples that can be allocated to this problem. 00:27:47.000 --> 00:27:52.000 But then it optimizes the allocation across the strata. 00:27:52.000 --> 00:27:58.000 Given that number of samples that are available to be taken. 00:27:58.000 --> 00:28:07.000 So the cost information the this module allows you to input the cost per sample and per analysis separately for each strata. 00:28:07.000 --> 00:28:17.000 I stratum. And so that makes sense because it could be the case that you're maybe your collection costs might be greater in zone, a due to Greater radiological controls or ppe. 00:28:17.000 --> 00:28:27.000 And so, and allowing those to be input separately will allow you to optimize the cost on the sample size more effectively. 00:28:27.000 --> 00:28:31.000 Taking that into account. 00:28:31.000 --> 00:28:36.000 So here, this just shows an example of how Vsp does that optimization? 00:28:36.000 --> 00:28:44.000 So in this first example here, it's the collection cost per sample was exactly the same for zone A and zone B. 00:28:44.000 --> 00:28:47.000 It takes those total of 65 samples and allocates 14 in zone a and 51 in zone. 00:28:47.000 --> 00:28:52.000 B, so that actually is a higher sample, density in zone a, since it is so much smaller. 00:28:52.000 --> 00:29:02.000 But that's so. That's driven slowly by the higher standard deviation of the higher variability in sound. 00:29:02.000 --> 00:29:19.000 A. And in this kind of alternate scenario here, suppose the collection cost per sample is actually 6 times as high in zone A as zone B, that just slightly reduces the number of samples allocated to zone a and bumps up the number of samples in zone 00:29:19.000 --> 00:29:32.000 B, and you do end up with a end up with a slightly higher cost just due to that really high collection costs there. But it comes up with a different allocation based off based off of those costs. 00:29:32.000 --> 00:29:38.000 So here again, just notionally, this kind of shows the type of sampling plan you might get from this kind of design. 00:29:37.000 --> 00:29:53.000 A, and kind of more spread out in Zone B, and if we were to collect the data and get results at each of those locations, you could then plot that in Vsp. 00:29:53.000 --> 00:30:01.000 And view those view those data points, and so you can kind of see here, there's definitely higher variability. 00:30:01.000 --> 00:30:11.000 Probably in this area, just by eyeballing it, maybe lower lower levels as well as lower variability in in zone B, so that data plotting. 00:30:11.000 --> 00:30:19.000 That's kind of like your first data quality check, just like looking at the data, seeing if it all makes sense and then Dsp can also perform data analysis to calculate the estimated mean. 00:30:19.000 --> 00:30:38.000 So the formula here is 2 is to combine all this, the straighta, as it takes into account the estimated need in the stratum as well as the proportion of the site in this sort of age, so this is an unbiased main estimate, and then similarly with standard error. 00:30:38.000 --> 00:30:55.000 for the standard deviation on that estimated sight mean that we were trying to control, for it calculates that in a way that accounts for those differing straighta size, and so what we come up with for on the test page for an estimate of the unbiased mean we can calculate 00:30:55.000 --> 00:31:04.000 the, sample mean of zone A and zone B, and then the estimated mean for the combined straighta is 2 57.5 3 pico caries program. 00:31:04.000 --> 00:31:12.000 And so what we see is if we were to just calculate just a simple, you know, basic average on those values, it would have been biased much higher that the mean would have been estimated at 402.2, so this helps us unbiased. 00:31:12.000 --> 00:31:26.000 Unbiased. That mean based off of the the straight on size. 00:31:25.000 --> 00:31:37.000 So the analysis of objective here is to create a specialistimate of the rateological contamination across both areas. 00:31:37.000 --> 00:31:44.000 Just using this data that we notionally got back from from this from the stratified sampling. 00:31:44.000 --> 00:31:54.000 And so we are going to use Geos statistical analysis for creaking to create an estimate map and delineate contours, and then look at, look at those delineated areas. 00:31:54.000 --> 00:32:03.000 So Fred talked about some pretty sophisticated geophysics and geo statistical analysis, which is a lot of where we wanna go with Esp. 00:32:03.000 --> 00:32:06.000 The current geostatistical analysis of Psp. 00:32:06.000 --> 00:32:17.000 We have designed to try to be pretty simple and pretty easy, so you can just kind of one click and get some SMS, but we do also give you a fair number of options to refine and improve these estimates. 00:32:17.000 --> 00:32:28.000 So we're always trying to strike that balance between making things simple to use and kind of oversimplifying things, or yeah, making it too much of a blockbox. 00:32:28.000 --> 00:32:28.000 So here. This shows the type of estimate you get backs. 00:32:28.000 --> 00:32:37.000 So this is this is using that notional data. This is kind of like the essence of the contamination. 00:32:37.000 --> 00:32:41.000 So you can see, you know. Obviously, we've got that kind of hot spot, high area. There. 00:32:41.000 --> 00:32:45.000 You know, we've got some of these other slightly elevated areas, and then you can use that estimate map to delineate contentors. 00:32:45.000 --> 00:32:48.000 And there's a fair number of different ways. You can do this here. 00:32:48.000 --> 00:32:53.000 I've just shown, you know, doing some kind of evenly spaced contours along the map. 00:32:53.000 --> 00:32:58.000 This can be helpful, for, like dilating remediation areas or exclusion zones, things like that. 00:32:58.000 --> 00:33:21.000 But and these can be. These can also be like exported to shapes, and used in that way to guide, maybe quite further sampling or guide division of the area into different decision units, for after after remediation. 00:33:21.000 --> 00:33:24.000 No, excuse me, another tool that you can use once you've done cring is this area delineation tool? 00:33:24.000 --> 00:33:36.000 So this actually kind of similar to the contour idea it identifies these elevated regions automatically from the create data, you can also define them manually. 00:33:36.000 --> 00:33:44.000 But what this allows you to do is probably a little hard to see with the small text here, but it basically gives you statistics on these delineated areas. 00:33:44.000 --> 00:33:44.000 So just looking at that subsystem of the site, what do I know about that site? 00:33:44.000 --> 00:33:58.000 What is the Craig estimate? Say about that site? So here this blue area I've gone through and done the automatic from creed data for anything over a 1,000, I believe. 00:33:58.000 --> 00:34:01.000 Yeah, it's a 1,000 Pico carries program. 00:34:01.000 --> 00:34:08.000 We were looking at, and so that it that once I've deliveated that it tells me, okay, this area is 120 acres. 00:34:08.000 --> 00:34:17.000 The average creature value is around 1322. There are only 7 measurements in that area, and then subsequent statistics on the actual measurements themselves. 00:34:12.000 --> 00:34:24.000 And then some statistics on the actual measurements themselves. So this can be helpful to kind of drill down into. 00:34:24.000 --> 00:34:26.000 Okay, what do I actually know about this region? And kind of do some do some analysis on that? 00:34:26.000 --> 00:34:30.000 So the bluery and the pink area. Then you can also compare those to background. 00:34:30.000 --> 00:34:33.000 So here, this is like kind of a box, and whisker plot of the cream values within those areas. 00:34:31.000 --> 00:34:39.000 Or the background. 00:34:39.000 --> 00:34:46.000 And then once those areas are, completed, we have a tool to estimate costs. 00:34:46.000 --> 00:34:49.000 And so this is based on the area size and parameters. 00:34:49.000 --> 00:34:54.000 So those those areas that I delineated have it have a specific size on the map. 00:34:54.000 --> 00:34:58.000 But then I can go in and specify a depth as well as a remediation cost in terms of there's a few different ways to input this. 00:34:58.000 --> 00:35:07.000 But, generally speaking, in terms of cost per volume, as well as the whatever fixed startup costs. 00:35:07.000 --> 00:35:12.000 So this can be a really useful tool to just kind of like estimate. Okay? 00:35:12.000 --> 00:35:16.000 Based on, based on the size of this area, based on what I expect my cost to be what? 00:35:16.000 --> 00:35:20.000 What would I expect the remediation cost to be? 00:35:20.000 --> 00:35:37.000 And you can get into a fair amount of detail with this, as far as like, you can define multiple layers within a particular remediation area, you know, if there's going to be different costs for the different layers or different portions and get fairly sophisticated with that with that cost analysis so in 00:35:37.000 --> 00:35:43.000 conclusion. So, yeah, I think we we've got a lot of interest in this area. 00:35:43.000 --> 00:35:47.000 I know we have a lot of ideas for further development. 00:35:46.000 --> 00:36:06.000 You know, these kind of this kind of statistical processes into characterization as well as some of these advanced and threed geophysical analysis capabilities and Vsp, and really kind of all geared towards providing support for integrated sites characterization all the way through the process you know that 00:36:06.000 --> 00:36:16.000 sample, design, analysis, and conceptual understanding, quantifying the uncertainty and repeating, repeating to the extent that is needed. 00:36:16.000 --> 00:36:19.000 And then that actually wraps things up as far as our site. 00:36:19.000 --> 00:36:25.000 So I think I think we open it up. We can open it up for questions. 00:36:25.000 --> 00:36:29.000 All right. Thank you. I see questions are starting to come in the room. 00:36:29.000 --> 00:36:31.000 So Bobby, go right ahead. Yeah, this is Bobby. 00:36:31.000 --> 00:36:38.000 It first of all, I want to thank Ben and Elstad. 00:36:38.000 --> 00:36:46.000 You're a great staff. They have lots of experience I do appreciate from an Nrc perspective to their support. 00:36:46.000 --> 00:36:51.000 To us, they do an outstanding job in their analysis. 00:36:51.000 --> 00:36:54.000 By name Lisa, Fred, and Jennifer as well. So thank you. 00:36:54.000 --> 00:37:06.000 So much for excellent presentation. How to say this is a continuing activity, that we are dealing with, because we had some issues make main issues for sub-surface characterization and sampling. 00:37:06.000 --> 00:37:13.000 We are trying to reduce the cost for sampling at the same time to make sure that we are dealing with risk. 00:37:13.000 --> 00:37:28.000 One area, the only mission that we're trying to add actually sampling and delaying the different layers or strata based on risk analysis. 00:37:28.000 --> 00:37:32.000 In order to establish these fgl for each layer. 00:37:32.000 --> 00:37:33.000 So it is very important work that we are doing. The statistical analysis standard deviation for concentration. 00:37:33.000 --> 00:37:52.000 This is definitely going to be like basis for an foundation for developing in the future what we call subsurface marsing, because margin currently falls surface. 00:37:52.000 --> 00:38:01.000 And this is definitely going to build the foundation for characterization and statistical approaches and sampling for suburface. 00:38:01.000 --> 00:38:07.000 Specifically when we talk about radioactive material for subsurface. Thanks again. 00:38:07.000 --> 00:38:10.000 Absolutely. Yeah. Thank you. Bobby. 00:38:10.000 --> 00:38:10.000 Thank you. 00:38:10.000 --> 00:38:12.000 Alright. Thank you. I do see a number of questions that came in online. 00:38:12.000 --> 00:38:26.000 So I'm going to lead in with them. First of all, a number of people have chimed in that they use Vsp, that it's been very helpful for them at their site. 00:38:26.000 --> 00:38:33.000 There are some requests. If you'll do a clue in Internet seminars on Vsp, so I may be following up with the 2 of you later about that one. 00:38:33.000 --> 00:38:41.000 But specifically, some have asked if Vsp can be used for air, quality, or Hindo air quality scenarios. 00:38:41.000 --> 00:38:44.000 Yeah, I think Eric quality is is a little tough. 00:38:44.000 --> 00:38:47.000 Just cause. Yeah, so vsp, is so driven by that conceptual site model. 00:38:47.000 --> 00:38:56.000 And I think probably a big, fundamental like assumption in Vsp, is that like when you sample, it's like representative of the population. 00:38:56.000 --> 00:38:57.000 So, and it is very spatially focused. So I think that's something that we're really looking at. 00:38:57.000 --> 00:39:06.000 So air and water. Both are kind of challenging, just because obviously, things are are just in flux. 00:39:06.000 --> 00:39:11.000 A little more. I believe I believe people have used Dsp. 00:39:11.000 --> 00:39:25.000 In terms of you know, if you are, if you're kind of if you're if you're kind of if your objective is like the detection like just, you know, detecting above the limit, Vsp can be used to kind of help with like air monitor like placement, but it does it does get a little challenging just 00:39:25.000 --> 00:39:30.000 because Vsp is so spatially focused. So I would say it kind of just it. 00:39:30.000 --> 00:39:37.000 Kind of just depends on whether that makes sense for what you're trying to sample for what they are something, and that's probably not a great answer. 00:39:37.000 --> 00:39:41.000 But I mean, I think that is kind of somewhat of a gap and something that we're definitely thinking about. 00:39:41.000 --> 00:39:43.000 All right. I have a question in reference to slide 32. 00:39:43.000 --> 00:39:49.000 They noted that you showed Google Earth plotting on slide. 00:39:49.000 --> 00:39:49.000 32, and they were just wondering what other spatial programs is. 00:39:49.000 --> 00:39:54.000 The output data from Vsp compatible with will it work with or is it compatible with, S. 00:39:54.000 --> 00:40:01.000 Re, or Qgis. 00:40:01.000 --> 00:40:13.000 Yeah, it should be. Yeah. So this actually is a screenshot of vsp, so here we're bringing in the like, kind of Google Earth stuff satellite imagery into Thesp, that's available directly you can do that. 00:40:13.000 --> 00:40:19.000 From Vsp. And I guess if folks are interested I could follow up with me later. 00:40:19.000 --> 00:40:23.000 And I can show you how to do that. But yeah, as far as export from Vsp. 00:40:23.000 --> 00:40:29.000 For sure. So vsp the site boundaries, as well as the shape files, or sorry the like map lines. 00:40:29.000 --> 00:40:31.000 You can export both of those to shapefiles. 00:40:31.000 --> 00:40:40.000 That's kind of a common way to get that out we you can also export, I believe, the samples to like go, Json, so there's a few different options there, but I think Shapefile is probably the most flexible for different software tools. 00:40:40.000 --> 00:40:49.000 To get it in. 00:40:49.000 --> 00:40:53.000 All right. Excellent! Leading in again with another online question for Krieuing. 00:40:53.000 --> 00:40:58.000 Can you overlay an error of estimate plot? 00:40:58.000 --> 00:41:05.000 Yes. Yeah. So, yeah. Vsp, when you do the cring, and I don't have a screenshot of it here. 00:41:05.000 --> 00:41:15.000 But because you can kinda can kind of see here, there's a creed estimates as well as the creaking variances, so yes, that can be displayed on the map as well. 00:41:15.000 --> 00:41:30.000 And actually got it. But there's a few different like probability and uncertainty. Maps that we do have in Vsp, so like Fred showed that probability of exceeding a threshold and indicator creating those are things that you can do with the Cring and Bsp as well. 00:41:30.000 --> 00:41:41.000 All right. And one other question online. This is about cost. How often is the cost estimate updated in the Vsp program? 00:41:41.000 --> 00:41:54.000 Yeah, that particular tool we have not updated the costs in quite a while, and I know people sometimes give us a hard time, like as far as some of our default costs are like really low. 00:41:54.000 --> 00:41:56.000 And so it's just kind of placeholder values the whole idea is that you should put in your own values for sure. 00:41:56.000 --> 00:42:03.000 But yeah, the we have not updated this particular cost module in a while. 00:42:03.000 --> 00:42:03.000 I think we'd be really interested if you use this, or if you're like, Hmm! 00:42:03.000 --> 00:42:08.000 This isn't quite useful, because it doesn't do this. 00:42:08.000 --> 00:42:16.000 Please let us know, because we'd love to hear about ways we could make it better or improve it. 00:42:16.000 --> 00:42:23.000 Alright excellent last chance for in-room questions. Oh, I see one go right ahead. 00:42:23.000 --> 00:42:34.000 So you you spoke to an iterative process. Is there anything with like grab samples versus composite, or any way to even cut costs even further? 00:42:34.000 --> 00:42:44.000 Where it starts starting off with maybe composite, and then homing in on areas that you can do a different type of gridding to do, grab, grab samples. 00:42:44.000 --> 00:42:56.000 Yeah, I think that's a great idea. We do there's there's a few different modules in Psp called multiple increments setpling, which is basically a type of composite sampling. 00:42:56.000 --> 00:42:58.000 And yeah, it's designed for that method of like, yeah, instead of like doing like individual samples. 00:42:58.000 --> 00:42:58.000 Your composite compositing them into your group. 00:42:58.000 --> 00:43:13.000 So there's methods to do that to support estimating the mean, comparing them into the threshold as well as like identifying elevated areas. 00:43:13.000 --> 00:43:16.000 And yeah, unfortunately, I didn't cover those in this slide. 00:43:16.000 --> 00:43:21.000 But but yeah, it is it allows you to place the samples in that way and then analyze them. 00:43:21.000 --> 00:43:25.000 Once you get your composites back. So yeah, there's some. 00:43:25.000 --> 00:43:30.000 There's some methods to do that. 00:43:30.000 --> 00:43:35.000 Alright. I see one more question in the room. 00:43:35.000 --> 00:43:41.000 Is there any way to customize your like, your sample layout, or your site itself? 00:43:41.000 --> 00:43:44.000 I'm just thinking back to what we've talked about this morning inorporating the environmental sequence, stratigraphy, taking that information where it kind of directs. 00:43:44.000 --> 00:44:07.000 Us. Let's say we have, like a Paleo valley, and you know you know your contamination is going to be flowing in this way instead of putting a box around in all of Canon Air Force Base and laying out a grid pattern sampling method could we can you customize this 00:44:07.000 --> 00:44:16.000 thing to focus in on just those areas for delineation and then get the same kind of statistical analysis from that kind of incorporating more of that. 00:44:16.000 --> 00:44:21.000 Geophysical algorithms to kind of direct your sampling. 00:44:21.000 --> 00:44:29.000 Yeah, I think you have absolutely can. And this is a large part of yeah, what we're kind of looking at as we move towards the at the, you know, the suburface problem. 00:44:29.000 --> 00:44:34.000 And things like that is getting a lot more sophisticated about the type of information you can brand to your conceptual site model. 00:44:34.000 --> 00:44:39.000 So yeah, and I think the the site that I've shown here is really pretty simplified. 00:44:39.000 --> 00:44:43.000 But you can get fairly sophisticated with Vsp, as far as like defining your survey. 00:44:43.000 --> 00:44:43.000 You know your decision unit, you know, just for the just for particular areas defining like holes or exclusion areas. 00:44:43.000 --> 00:45:00.000 So I mean I guess my answer would be that I think you can do that, but I think there's probably tools and things we could build in to make that easier to bring in that external information and really help it shape the sampling. 00:45:00.000 --> 00:45:11.000 So you can kind of do it manually in Vsp, right now. But I think there's a lot of room to make that better and bring that information in.